Mass spectral characterization of submicron biogenic organic particles in the Amazon Basin

Submicron atmospheric particles in the Amazon Basin were characterized by a high‐resolution aerosol mass spectrometer during the wet season of 2008. Patterns in the mass spectra closely resembled those of secondary‐organic‐aerosol (SOA) particles formed in environmental chambers from biogenic precursor gases. In contrast, mass spectral indicators of primary biological aerosol particles (PBAPs) were insignificant, suggesting that PBAPs contributed negligibly to the submicron fraction of particles during the period of study. For 40% of the measurement periods, the mass spectra indicate that in‐Basin biogenic SOA production was the dominant source of the submicron mass fraction, contrasted to other periods (30%) during which out‐of‐Basin organic‐carbon sources were significant on top of the baseline in‐Basin processes. The in‐Basin periods had an average organic‐particle loading of 0.6 μg m−3 and an average elemental oxygen‐to‐carbon (O:C) ratio of 0.42, compared to 0.9 μg m−3 and 0.49, respectively, during periods of out‐of‐Basin influence. On the basis of the data, we conclude that most of the organic material composing submicron particles over the Basin derived from biogenic SOA production, a finding that is consistent with microscopy observations made in a concurrent study. This source was augmented during some periods by aged organic material delivered by long‐range transport.


Introduction
[2] Organic material (OM) typically constitutes 70 to 90% of the fine mass fraction in the Amazon Basin [Fuzzi et al., 2007]. During the wet season, natural processes dominantly influence the loadings and the size distributions of the particles [Andreae et al., 2002;Martin et al., 2009]. These processes include the emission of PBAPs, the production of biogenic SOA, and the episodic input of marine and African particles by long-range transport. The finefraction organic-carbon (OC) loading is greater in the day than at night, perhaps because of enhanced SOA production by the photochemical oxidation of biogenic precursor gases [Graham et al., 2003]. Speciation studies by chromatography show the presence in the fine fraction of methyltetrols, which are produced by the oxidation of isoprene [Claeys et al., 2004]. Quantitative characterization of different OC sources in the Amazon Basin, however, is limited and challenging. The contribution of PBAPs such as bacteria and spores to the fine fraction is poorly understood and quantified [Elbert et al., 2007]. The influence of downward convective mixing of particles from aloft (e.g., delivered by long-range transport) remains uncertain. These uncertainties result in loose constraints and hence limited accuracy for the modeling of scattering and absorption of solar radiation by submicron particles over the Amazon Basin and their activity as cloud condensation nuclei (CCN).
[3] In the wet season of 2008, as a part of the Amazonian Aerosol Characterization Experiment (AMAZE-08), we deployed a high-resolution time-of-flight aerosol mass spectrometer (AMS). This instrument conducts real-time measurements of the size-resolved chemical composition of non-refractory submicron particles [DeCarlo et al., 2006]. Herein, we investigate the use of the mass spectra for characterizing the sources and properties of Amazonian particles.

Experiment
[4] Ground-based measurements were carried out from 7 Feb to 13 Mar 2008. The rain forest site (02°35.68 0 S, 60°1 2.56 0 W, 110 m above sea level), located 60 km NNW of Manaus, faced ca. 1600 km of nearly pristine forest to the east. The ten-day back trajectories indicated arrival of air masses from the northeast, originating over the Atlantic Ocean in the direction of Cape Verde and the Canary Islands. Aerosol from above the forest canopy (33 m) was sampled from the top of a tower (''TT34''; 38.75 m). The meteorological parameters at the tower top were 296/298 K (25 and 75% quantiles), 89/100% relative humidity (RH), and 995 ± 5 hPa during the measurement period. The aerosol was drawn to the ground in a turbulent flow of 40 to 80 L min À1 through a stainless steel tube (1.27 cm outside diameter), was passed through a Nafion dryer (exit RH varied from 40% to 70%), and then was sampled by the AMS. Most of the results in this paper pertain to the data set collected using this AMS, but a second AMS operated by MPI-C sampled on a laminar-flow line (20 to 35% RH). The major results reported in this paper were confirmed by co-analysis of the MPI-C data set.
[5] The AMS data were processed by standard methods (see auxiliary material). 1 The mass loadings were adjusted to standard temperature and pressure (STP; 273 K and 10 5 Pa), which were approximately 10% greater than those at ambient conditions. A value of 1.0 was used for the AMS collection efficiency (a factor which potentially corrects for undetected particle mass) on the basis of the observation of liquid-like organic particles by microscopy [Sinha et al., 2009]. With this collection efficiency, the campaign-average sulfate loading measured by the AMS (0.15 ± 0.03 mg m À3 ; <1 mm) agreed within uncertainty with measurements made by ion chromatography (0.21 ± 0.06 mg m À3 ; <2 mm) and particle-induced X-ray emission (0.24 ± 0.05 mg m À3 ; <2 mm) for filter samples. For comparison, the average sulfate loading measured in four previous campaigns ranged from 0.17 to 0.26 mg m À3 in the wet season (cf. Martin et al., 2009, Table 3). The campaign-average OM loading measured by the AMS was 0.7 ± 0.3 mg m À3 (<1 mm). This loading is at the low end of the range of 1.0 to 2.0 mg m À3 (<2.5 mm; OM:OC = 1.7) reported in the previous campaigns [Fuzzi et al., 2007] (see auxiliary material for further explanation).
[6] For the analysis herein, 12-h mean daytime (06:00 to 18:00) and nighttime (18:00 to 06:00) mass loadings were calculated from the AMS data. Periods having low data coverage during a 12-h averaging time were removed. Measurement periods that were influenced by changes in local winds that could have caused the exhaust of the on-site power generator or regional anthropogenic outflow from Manaus to arrive at the tower were also excluded. Elemental oxygen-to-carbon (O:C) and hydrogen-to-carbon (H:C) ratios were calculated using 12-h high-resolution mass spectra [Aiken et al., 2008].

Results and Discussion
[7] Correlative behavior between submicron organic and sulfate loadings, given the greater existing knowledge concerning the processes controlling sulfate and its precursors in the Amazon Basin, can give insight into the processes that influence the organic content. Sulfate is the predominant inorganic constituent of wet-season Amazonian particles [Fuzzi et al., 2007]. Approximately 0.05 mg m À3 of the sulfate loading (<2 mm) has been attributed to in-Basin biogenic sources derived from sulfur-containing gases (e.g., DMS and H 2 S) emitted from the rain forest ecosystem [Andreae et al., 1990]. Sulfate loading above the ecosystem baseline occurs by out-of-Basin advective input of both marine-derived sulfate from the Atlantic Ocean and combustion-derived sulfate from Africa.
[8] During AMAZE-08, organic and sulfate loadings are well correlated within classes ( Figure 1a). A histogram of sulfate mass fraction shows three modes ( Figure S1), and consequently three classes I, II, and III of data points are colored in Figure 1 for low, moderate, and high sulfate mass fractions. The stratification by sulfate mass fraction further categorizes well the mass spectra of the organic material ( Figure 1b). The ratio of the signal intensity at m/z 44 to that at m/z 43, which are the most intense in the organic mass spectra, is constant within uncertainty for each class. An elevated 44-to-43 ratio has been suggested as an indicator of more oxidized organic material [Shilling et al., 2009]. As a group, the organic mass spectra within class I are selfsimilar and likewise within class III. Self-similarity is  defined by the criterion that the standard deviations of the relative intensities of prominent peaks are mostly within 15%, although there are a few exceptions (cf. Table S1). The spectra of class II can be divided into two sub-classes, one of which is similar to that of class I and the other to that of class III.
[9] These classifications can be associated with sources of the organic material. For instance, in-Basin biogenic SOA production can be concluded as the dominant source of submicron material during class I periods. The mass spectra (Figure 2a) have similar major peaks as those of laboratory SOA particles formed from the oxidation of individual biogenic precursors [Bahreini et al., 2005;Shilling et al., 2009]. High-resolution mass spectra of SOA particles, obtained in the Harvard Environmental Chamber (unpublished data) for the oxidation of isoprene, the monoterpene a-pinene, and the sesquiterpene b-caryophyllene, can be linearly combined to largely reproduce the patterns observed in AMAZE-08. The O:C ratio, which ranged from 0.39 to 0.45 (avg. 0.42; OM:OC % 1.71) during class I periods, is also similar to that of laboratory SOA particles at realistically-low precursor concentrations (e.g., 0.40 to 0.45 by Shilling et al. [2009]). For comparison, urbancombustion primary emissions (O:C < 0.10) and aged regional particles (O:C % 0.9) observed nearby Mexico City had significantly different values [Aiken et al., 2008]. The conclusion is that, even though the molecular details are expected to have differences owing to the greater complexity of Amazonian production pathways (e.g., emissions and reactions) compared to laboratory processes, overall properties like the O:C ratio can still be similar. The explanation may be that volatility influences the partitioning between gas and particle phases and that a predictive factor of volatility can be the O:C ratio within many molecular classes.
[10] Some other plausible alternatives to an in-Basin SOA origin for class I particles, such as regional biomass burning, emissions of PBAPs, or an out-of-Basin influence, can be considered but ruled out. The mean sulfate loadings were below 0.1 mg m 3 , a threshold value that is conservatively consistent with in-Basin sources (i.e., twice 0.05 mg m À3 taking into account uncertainty), thereby ruling out a significant out-of-Basin influence. The relative intensity of m/z 60, which is a fragment of levoglucosan and related species and serves as a marker of biomass burning, was less than 0.3% during AMAZE-08 and therefore indicating the absence of a significant influence of regional emissions from biomass burning [Docherty et al., 2008]. In support, measurements using a PTR-MS during AMAZE-08 recorded benzene-to-acetonitrile ratios varying from 0.0 to 0.3, effectively setting a ten-day lower limit based on photochemical aging since the last exposure of an air mass to tropical biomass burning (i.e., a typical ratio of 1.0 at emission [Karl et al., 2007]).
[11] Comparison of the mass spectra for model PBAPs (J. Schneider, unpublished data) to those for AMAZE-08 shows that characteristic peaks, such as m/z 60, 61, and 73 for carbohydrates like sugar alcohols and m/z 30, 42, and 56 for amino acids, constitute no more than 5% of the organic material during AMAZE-08. Furthermore, nitrogencontaining peaks expected for amino acids are negligibly present in the high-resolution mass spectra (cf. Figures 2a  and 2b). Collaborators in AMAZE-08 imaged submicron particles collected on filters and observed that spherical organic particles that appear as liquid droplets are the main population [Sinha et al., 2009], a result which is consistent with a dominant presence of biogenic SOA particles and a minor contribution of PBAPs to the submicron fraction.
[12] In contrast to class I periods, elevated mean sulfate loadings (i.e., 0.2 to 0.6 mg m À3 ) suggest a significant contribution from out-of-Basin sources during class III periods. This contribution is further supported by the similarity of a representative organic mass spectrum (Figure 2b) to those of aged rural particles [Zhang et al., 2005] and freetroposphere particles [Hock et al., 2008]. A difference mass spectrum between class I and III highlights an elevated relative intensity at m/z 44 (Figure 2c), suggesting a contribution by highly oxidized materials [Zhang et al., 2005]. The overall O:C ratio correspondingly increased during class III periods, with a range of 0.46 to 0.53 (avg. 0.49; OM:OC % 1.79). Good candidates for the sources of these oxidized materials include emissions from African biomass burning [see Capes et al., 2008, Figure 9d] and possibly the Atlantic Ocean, followed by in-and out-ofcloud photo-and oxidant processing during transport.
[13] Diel profiles shown in Figure 3 for AMAZE-08 are consistent with the above descriptions. The surface loadings families, as obtained by analysis of the high-resolution mass spectra. The shown mass-to-charge (m/z) ratios are truncated at 100, and the higher m/z values contribute no more than 5% to the total organic signal. Inset pie charts represent the period-average chemical composition. (c) A difference plot of the class I and class III spectra shown in Figures 2a and 2b. in the morning are influenced by the residual layer of the previous day (i.e., formed above the nocturnal boundary), day-to-day variability of boundary layer depth, and nighttime rainfall. For class I periods (Figure 3a), a doubling of the organic loading from the morning to the afternoon and a decline from the late afternoon to next morning are suggestive of photochemically initiated emissions and SOA production. The maximum (16:00; Figure 3a) is consistent with an earlier increase in the concentrations of isoprene and monoterpenes (13:00-15:00; Figure 3c), which are SOA precursor gases. The diel invariance of the m/z 44-to-43 ratio suggests a single source ( Figure S2a), and its value (ca. 1.0) indicates little atmospheric aging. In comparison, for class III periods the m/z 44-to-43 ratio increases from approximately 1.3 to 2.0 in the morning and then maintains a steady state ( Figure S2b), supporting the explanation of daytime convective downward mixing of aged organic particles from aloft. Consistent with this explanation, the organic loading increases more in the morning during class III than class I periods and then maintains an apparent steady state for the rest of the day (Figure 3b). The sulfate loading shows a similar but even stronger diel trend, providing further indication to the importance of long-range transport during class III periods ( Figure S2d).
[14] In contrast to the clear stratifications of mean sulfate loadings of <0.1 and >0.1 mg m À3 during class I and III periods, respectively, the sulfate loading ranges from 0.06 to 0.47 mg m À3 during class II periods. The organic mass spectra are also better represented by two bifurcating branches, one closer to the spectra of class I and the other to that of class III, rather than as an average. The O:C ratios range from 0.37 to 0.55. Class II periods can therefore be explained as mixed periods during which the out-of-Basin influence is significant yet not fully emergent. For the period of study, we classify 40% of the time as dominantly affected by in-Basin sources, 30% as strongly influenced by out-of-Basin sources on top of the baseline in-Basin sources, and the remainder as having a significant but weaker out-of-Basin influence.
[15] The measurements reported herein can be put into greater context by comparison with the predictions of a chemical transport model. GEOS-Chem was employed (see auxiliary material) to simulate the loadings at the site during the time period of AMAZE-08. The simulated average organic loading was 35% lower than the measurement. This underprediction is several-fold smaller than simulationmeasurement comparisons made for anthropogenically influenced regions [e.g., Volkamer et al., 2006]. Although the reduction in the underprediction for the natural conditions of AMAZE-08 is encouraging as a metric for confidence in the accuracy and completeness of the underlying physics and chemistry of the simulation, the possibility of fortuitous cancelling of errors must be considered [Capes et al., 2009]. Many uncertainties remain in the simulation, including the identification and emission rates of biogenic SOA precursors, the formation mechanisms (e.g., heterogeneous and aqueous processes), the accuracy of the included oxidant radical cycle to the Amazonian boundary layer [Lelieveld et al., 2008], and the SOA particle mass yields [Shilling et al., 2009].
[16] In summary, this paper presents the first real-time mass spectrometric measurements of the chemical composition of submicron Amazonian particles. The particles were largely composed of sulfate and organic material, having a campaign-average organic mass fraction of 0.78. The results further show that biogenic SOA production can dominate the submicron organic loading during times of diminished out-of-Basin influence. SOA production is expected to occur both in the gas and particle phases and in cloud waters, and this production is expected to strongly influence the size distribution and the chemical composition of submicron particles in the Amazon Basin, in turn regulating the CCN and optical properties of the particle population. During time periods of strong out-of-Basin influence, the organic loading increased on average from 0.6 to 0.9 mg m À3 (i.e., +50%). The material was also more oxidized, and this altered composition might be hypothesized to influence the physical or reactive properties of the particles.
[17] Acknowledgments. Support was received from the U.S. National Science Foundation, the German Max Planck Society, and Brazilian CNPq and FAPESP agencies. QC acknowledges a NASA Earth and Space Science Fellowship. DKF acknowledges a NOAA Global Change Fellowship.  [Karl et al., 2009].